RFR JDK-8194084: Obsolete FastTLABRefill and remove the related code

[Stuart Monteith]

Hi,
   JC's patch looks good. This provides an explanation for the
spurious ldp I was  . I'm currently building aarch64/x86_64, and will
run some tests.

Thanks!
    Stuart

On 20 February 2018 at 16:33, Andrew Dinn <adinn at redhat.com> wrote:
> Hi JC,
>
> On 20/02/18 16:22, JC Beyler wrote:
>> I fixed this issue after Stuart + Andrew diagnosed the issue for
>> aarch64. After looking at the code and trying to get the spill/fills out
>> of the way when possible, x86 also had the possibility to do skip a case
>> of spill/fill.
>>
>> Let me know what you think:
>> http://cr.openjdk.java.net/~jcbeyler/8194084/webrev.05
>>
>> I added a no_pop slowpath, fixed the str to stp for aarch64, and then
>> did the same no_pop slowpath for x86. I can also remove the no_pop label
>> and just fix the aarch64 code by moving the str/ldr to stp/ldp and
>> moving the stp before the jump to the slowpath.
>>
>> Let me know and thanks!
> That patch is fine for AArch64 and with it the build now completes and
> runs ok.
>
> I have not tested the change on x86 -- it looks fine by eyeball but it
> really needs a proper check.
>
> I assume you will raise a separate JIRA for this new patch?
>
> regards,
>
>
> Andrew Dinn
> -----------
> Senior Principal Software Engineer
> Red Hat UK Ltd
> Registered in England and Wales under Company Registration No. 03798903
> Directors: Michael Cunningham, Michael ("Mike") O'Neill, Eric Shander